RGBD Tracking Benchmark Dataset
收藏arXiv2012-12-12 更新2024-07-31 收录
下载链接：
http://occlusion.In
下载链接
链接失效反馈官方服务：
资源简介：
RGBD Tracking Benchmark Dataset是由香港科技大学和麻省理工学院合作创建的一个包含100个RGBD视频的数据集，旨在为跟踪算法提供一个高多样性的基准。该数据集涵盖了可变形物体、多种遮挡条件和移动相机，以及不同的光照和场景条件。数据集的创建过程中，手动标注了所有帧的地面实况，以确保高一致性。该数据集主要应用于计算机视觉领域，特别是解决跟踪任务中的模型漂移和遮挡问题，通过深度信息提高跟踪性能。
The RGBD Tracking Benchmark Dataset is a curated dataset consisting of 100 RGBD videos co-developed by The Hong Kong University of Science and Technology and the Massachusetts Institute of Technology, which aims to provide a high-diversity benchmark for tracking algorithms. This dataset covers deformable objects, diverse occlusion scenarios, moving cameras, as well as varying lighting and scene conditions. During its construction, all frames were manually annotated with ground truth to ensure high annotation consistency. It is primarily applied in the field of computer vision, specifically to address model drift and occlusion issues in tracking tasks, and to improve tracking performance by utilizing depth information.
提供机构：
香港科技大学
创建时间：
2012-12-12
